New Logo Pipeline Topology — Cross-System with Attribution Correction

FY26 closed won — outbound

0.0%

227 leads (corrected) · 0 won

Strict LEADSOURCE overwrites detected

2,090

External System → Outreach dominant: 1,388

FY26 outbound real-owner ≤24h

3.96%

9 of 227 leads · 96% sit with service users / routing queues

FY26 outbound engagement → won gap

85% → 0%

194 engaged · 0 won · conversion problem, not engagement

3D TOPOLOGY · drag to orbit · stage progression L→R · click axis labels for definitions

W ← stage progression →

B ← segment →

Z ← stuckness depth →

Hover surface · click heatmap labels

SURFACE COLOR

Healthy flow

Pool forming

Deep pool (trap)

PATHS BY SEGMENT

Outbound

Inbound

Webinar/Event

Product-Led

Untracked

Other (External System blob)

PROSPECT_TYPE OVERLAY · % lead-typed (vs contact-typed) at each stage · current cohort: FY26 · ⓘ how to read this SFDC Lead SFDC Contact

2D Heatmap · % cohort reaching each stage · all 6 segments × 9 stages · click axis labels for definitions

FY20

50,788 leads · pre-Outreach era

FY21

10,981 leads · volume cliff

FY22

103,674 leads · PLG peak (75,555 Product-Led)

FY23

65,288 leads · PLG → outbound transition

FY24

15,285 leads · outbound build (6,508 NNL)

FY25

9,021 leads · outbound collapse

FY26 (corrected)

2,812 leads · 1,388 reattributed

IDEAL STATE

~2,800 leads · attribution fixed, mix rebalanced, healthy rates · ~71 won

SEGMENT GLOSSARY

how each segment maps to Salesforce LEADSOURCE · FY26 → ideal

Outbound 227 → 800

LEADSOURCE = Outreach

Cold prospects pushed into an SDR sequence. SDR-driven, not self-identified.

Inbound 853 → 1,200

LEADSOURCE ∈ Website Form, HubSpot Chat, External System, Community, Inbound Email

Lead raised their hand to marketing on their own.

Webinar/Event 181 → 400

LEADSOURCE ∈ Webinar, Product/Sales Led Event

Captured from event registration or post-event follow-up.

Product-Led 22 → 300

LEADSOURCE = Dashboard User Creation

Free Cloud signup created the lead. FY22→FY26 collapse is mostly a tracking taxonomy change.

Untracked 114 → 0

LEADSOURCE IS NULL or unmappable

Source field never populated. Goal in ideal state is zero.

Other 1,415 → 100

LEADSOURCE = External System (post-overwrite-reversal)

Integration noise — ZoomInfo, content-gate triggers, third-party pushes. Where attribution corruption is most visible.

THREE ERAS · FY20 → FY26

PLG era (FY20–FY23): Product-Led (Dashboard User Creation) was the dominant entry point. FY22 peaked at 75,555 Product-Led leads (73% of cohort). Outbound essentially zero.

Outbound build-out (FY24): Outbound spikes from 420 (FY23) to 6,508 (FY24) — strategic motion shift to cold sequencing. Product-Led collapses to 65 (taxonomy change, not real decay).

Outbound collapse + attribution corruption (FY25–FY26): Outbound shrinks (479 → 227); what remains gets LEADSOURCE-overwritten by sequence entry. 2,090 strict overwrites detected total (1,409 in FY26 alone); 1,388 of all-time overwrites follow the External System → Outreach pattern.

⚠ The FY22→FY24 Product-Led cliff (75,555 → 65) is largely a tracking taxonomy change, NOT a real funnel collapse. Decay in the visualization mixes real GTM motion change with upstream definition shifts.

Leak Migration · FY20 → FY26 + IDEAL · where the biggest drop landed each year year-over-year leak position FY26 leak IDEAL benchmark

Where leads go to nowhere · attribution corruption is the lead story

ATTRIBUTION CORRUPTION · CHRONIC

2,090 leads had their LEADSOURCE overwritten — 1,388 were inbound traffic mis-tagged as Outbound at sequence entry.

Snapshot SCD2 detected 12,574 lead source changes; 2,090 are strict overwrites (a real first-touch source clobbered, not enrichment from null). The dominant pattern: External System → Outreach (1,388). Inbound integration leads (ZoomInfo, content gates, third-party trigger systems) get sequenced by SDRs, and the act of adding to a sequence overwrites the source field. In FY26, this means 86% of what was reported as "Outbound" was never actually outbound — it was inbound traffic with stolen attribution.

Strict overwrites: 2,090 · External → Outreach: 1,388 · Webinar → Outreach: 37

OWNERSHIP + CONVERSION · BOTH BROKEN

Real outbound sequences fast and engages well — then dies. 4% real owner, 0% closed won.

After stripping the 1,358 mis-tagged inbound leads from FY26 Outbound, the remaining 227 real outbound leads sequence in hours (median 0.01 days) and engage at 85% — that part works. The breakdown is downstream: real-human-owner-≤24h is only 3.96% (9 of 227); the other 96% sit with service users or routing queues. Of the 194 that engaged, 14 booked a meeting (6.17%), 6 opened an opp (2.64%), 0 closed won. The 35-day sequenced lag and 0.76% real-owner numbers from earlier analyses were artifacts of the inbound-mistakes; once those are removed, the real story is engagement without conversion.

FY26 outbound (corrected): 3.96% real owner · 85% engage · 6.17% meeting · 0% won

MARKETING-SALES HANDOFF · BROKEN

0% of outbound leads are reaching MQL. By design, but it's the gap the agentic-pipeline plan must close.

HubSpot lifecycle stage 03 (MQL) is reached by 0% of FY26 outbound leads — but FY22 was 63%, FY25 was 63%, FY20–FY23 ranged 7–63%. Only FY26 is 0%. Outbound has gone from connected-to-marketing-scoring to completely bypassed in one year — a new failure pattern, not a structural one. Tim's agentic-pipeline directive is exactly this fix: build a Snowflake Cortex score that fuses product signals + intent data + behavioral signals into a routable score, then use ChiliPiper Distro / SFDC routing to assign real owners — replacing the SVC Integrations black hole. (HubSpot's MQL scoring layer is being unwound as part of the same migration.)

FY26 outbound MQL reach: 0% · stage 04 (SQL) unused (n=1 in 12mo per Sarah's data)

Methodology · sourcing · caveats

Path: ANNA Thread 019ddf54 joined RAW.STITCH_SALESFORCE + RAW.HOCKEYSTACK + HUBSPOT_DATASHARE + RAW.OUTREACH_DATA + RAW.CHILIPIPER_DATA + ANALYTICS.DBT.MART_PROSPECTS + SNAPSHOTS.SALESFORCE.LEAD (SCD2). 257,849 new-logo leads across FY20–FY26.

Stage taxonomy: Aligned to Sarah's MART_PROSPECTS HubSpot lifecycle (Subscriber 01 → Lead 02 → MQL 03 → Opp Opened 05 → Won 16 / Paid 17). Stage 04 (SQL) skipped — Sarah's data shows n=1 in 12 months. Web Visit dropped (HockeyStack methodology artifact — fires on Outreach email clicks, structurally misleading for outbound segments). Subscriber inverted to "New Logo" = 100 − Subscriber reach (% of segment that is not yet a HubSpot subscriber). Click any stage label in the heatmap for full definition.

Real-owner rule: USER.ISACTIVE=TRUE AND USERTYPE='Standard' AND TERMINATION_DATE__C IS NULL/future AND user is NOT SVC Integrations or any *Integrations service user. SVC Integrations is explicitly excluded (75% of all rows in MART_PROSPECTS — the dominant data-hygiene gap on prior cuts).

FY26 = corrected: The FY26 cohort applies strict LEADSOURCE overwrite reversal. The "Other" segment in FY26 (1,415 leads) is the visible attribution-corruption surface — most originated as External System and were clobbered to Outreach at sequence entry; not heuristically reclassified into Inbound.

HockeyStack coverage gap (resolved): data ended 2025-12-15, leaving Web Visit null for FY26 cohorts. Web Visit is now dropped from the funnel (the email-click contamination made it noise anyway), so the gap no longer affects the visualization.

ChiliPiper coverage gap: data starts 2025-09-16. For FY24 and FY25 Jan–Aug, stage 8 (Meeting Booked) uses an SFDC Task/Event heuristic. The `meeting_source_breakdown` field per cell distinguishes ChiliPiper vs heuristic counts.

Outreach attribution path: Outreach prospects don't carry SFDC IDs — joined to Lead via email only. Sequence enrollment within 90d before lead_created_at filtered out.

Owner-24h is a proxy: No LEADHISTORY raw table; LEAD.SYSTEMMODSTAMP within 24h of CREATEDDATE is used instead.

10x test: FY24 outbound actuals overlaid on FY26 base. Answers "if we restart cold outbound at FY24 volume on today's pipeline, what does it look like?"

Cross-system pipeline, attribution-corrected.

Methodology · sourcing · caveats